Lung Segmentation in Feature Images with Gray and Shape Information
2014 ◽
Vol 513-517
◽
pp. 3069-3072
◽
Keyword(s):
The Mean
◽
Accurate lung segmentation in chest radiography is an important and difficult task in the development of computer-aided diagnosis. Therefore, we proposed a lung segmentation method in feature images with gray and shape information. Firstly, we extracted six feature images, and built an initial shape model. Then, we calculated the gray cost in the feature images. Finally, the lung profile was determined by use of shape restriction. With the feature images and method of shape restriction, the mean overlap rate was improved to 75.60%. Therefore, the method proposed in our study can improve the performance of lung segmentation.